Support GIN Indexes with ORCA.

This commit adds the GPDB side changes required to support GIN Indexes with
ORCA.  It also adds a new test file gp_gin_indexes to test plans produced for
ORCA/planner.

GIN indexes are not supported with index expression or predicate constraints.
ORCA does not support it currently for other types of indexes too.

src/backend/cdb/cdbpartindex.c

@@ -587,7 +587,11 @@ recordIndexesOnLeafPart(PartitionIndexNode **pNodePtr,
 		 */
 		indRel = index_open(indexoid, NoLock);
 
-		if (GIST_AM_OID == indRel->rd_rel->relam)
+		if (GIN_AM_OID == indRel->rd_rel->relam)
+		{
+			indType = INDTYPE_GIN;
+		}
+		else if (GIST_AM_OID == indRel->rd_rel->relam)
 		{
 			indType = INDTYPE_GIST;
 		}
@@ -1553,7 +1557,11 @@ logicalIndexInfoForIndexOid(Oid rootOid, Oid indexOid)
 	}
 
 	plogicalIndexInfo->indType = INDTYPE_BITMAP;
-	if (GIST_AM_OID == indRel->rd_rel->relam)
+	if (GIN_AM_OID == indRel->rd_rel->relam)
+	{
+		plogicalIndexInfo->indType = INDTYPE_GIN;
+	}
+	else if (GIST_AM_OID == indRel->rd_rel->relam)
 	{
 		plogicalIndexInfo->indType = INDTYPE_GIST;
 	}

src/backend/gpopt/translate/CTranslatorRelcacheToDXL.cpp

@@ -1116,7 +1116,12 @@ CTranslatorRelcacheToDXL::RetrieveIndex
 	
 		index_type = IMDIndex::EmdindBtree;
 		IMDRelation::Erelstoragetype rel_storage_type = md_rel->RetrieveRelStorageType();
-		if (GIST_AM_OID == index_rel->rd_rel->relam)
+		if (GIN_AM_OID == index_rel->rd_rel->relam)
+		{
+				index_type = IMDIndex::EmdindGin;
+				mdid_item_type = GPOS_NEW(mp) CMDIdGPDB(GPDB_ANY);
+		}
+		else if (GIST_AM_OID == index_rel->rd_rel->relam)
 		{
 			index_type = IMDIndex::EmdindGist;
 			mdid_item_type = GPOS_NEW(mp) CMDIdGPDB(GPDB_ANY);
@@ -1381,7 +1386,7 @@ CTranslatorRelcacheToDXL::RetrievePartTableIndex
 	default_levels_derived->Release();
 	mdid_index->AddRef();
 
-	GPOS_ASSERT(INDTYPE_BITMAP == index_info->indType || INDTYPE_BTREE == index_info->indType || INDTYPE_GIST == index_info->indType);
+	GPOS_ASSERT(INDTYPE_BITMAP == index_info->indType || INDTYPE_BTREE == index_info->indType || INDTYPE_GIST == index_info->indType || INDTYPE_GIN == index_info->indType);
 
 	IMDIndex::EmdindexType index_type = IMDIndex::EmdindBtree;
 	IMDId *mdid_item_type = NULL;
@@ -1393,6 +1398,11 @@ CTranslatorRelcacheToDXL::RetrievePartTableIndex
 	else if (INDTYPE_GIST == index_info->indType)
 	{
 		index_type = IMDIndex::EmdindGist;
+		 mdid_item_type = GPOS_NEW(mp) CMDIdGPDB(GPDB_ANY);
+	}
+	else if (INDTYPE_GIN == index_info->indType)
+	{
+		index_type = IMDIndex::EmdindGin;
 		mdid_item_type = GPOS_NEW(mp) CMDIdGPDB(GPDB_ANY);
 	}
 
@@ -3394,7 +3404,8 @@ CTranslatorRelcacheToDXL::IsIndexSupported
 	// index expressions and index constraints not supported
 	return gpdb::HeapAttIsNull(tup, Anum_pg_index_indexprs) &&
 		gpdb::HeapAttIsNull(tup, Anum_pg_index_indpred) &&
-		(BTREE_AM_OID == index_rel->rd_rel->relam || BITMAP_AM_OID == index_rel->rd_rel->relam || GIST_AM_OID == index_rel->rd_rel->relam);
+		(BTREE_AM_OID == index_rel->rd_rel->relam || BITMAP_AM_OID == index_rel->rd_rel->relam || GIST_AM_OID == index_rel->rd_rel->relam ||
+		 	GIN_AM_OID == index_rel->rd_rel->relam);
 }
 
 //---------------------------------------------------------------------------

src/include/nodes/plannodes.h

@@ -488,7 +488,8 @@ typedef enum LogicalIndexType
 {
 	INDTYPE_BTREE = 0,
 	INDTYPE_BITMAP = 1,
-	INDTYPE_GIST = 2
+	INDTYPE_GIST = 2,
+	INDTYPE_GIN = 3
 } LogicalIndexType;
 
 typedef struct LogicalIndexInfo

src/test/regress/parallel_schedule

@@ -55,6 +55,8 @@ test: copy copyselect
 test: create_misc create_operator
 # These depend on the above two
 test: create_index create_view
+# Depends on things setup for create_index
+test: gp_gin_index
 
 test: inherit
 

src/test/regress/sql/gp_gin_index.sql

+CREATE INDEX jidx ON testjsonb USING gin (j);
+SET optimizer_enable_tablescan = off;
+SET enable_seqscan = off;
+set enable_bitmapscan = on;
+
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"wait":null}';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"wait":"CC"}';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"wait":"CC", "public":true}';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"age":25}';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"age":25.0}';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"array":["foo"]}';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"array":["bar"]}';
+
+SELECT count(*) FROM testjsonb WHERE j @> '{"wait":null}';
+SELECT count(*) FROM testjsonb WHERE j @> '{"wait":"CC"}';
+SELECT count(*) FROM testjsonb WHERE j @> '{"wait":"CC", "public":true}';
+SELECT count(*) FROM testjsonb WHERE j @> '{"age":25}';
+SELECT count(*) FROM testjsonb WHERE j @> '{"age":25.0}';
+SELECT count(*) FROM testjsonb WHERE j @> '{"array":["foo"]}';
+SELECT count(*) FROM testjsonb WHERE j @> '{"array":["bar"]}';
+-- exercise GIN_SEARCH_MODE_ALL
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{}';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j ? 'public';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j ? 'bar';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j ?| ARRAY['public','disabled'];
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j ?& ARRAY['public','disabled'];
+
+SELECT count(*) FROM testjsonb WHERE j @> '{}';
+SELECT count(*) FROM testjsonb WHERE j ? 'public';
+SELECT count(*) FROM testjsonb WHERE j ? 'bar';
+SELECT count(*) FROM testjsonb WHERE j ?| ARRAY['public','disabled'];
+SELECT count(*) FROM testjsonb WHERE j ?& ARRAY['public','disabled'];
+
+-- array exists - array elements should behave as keys (for GIN index scans too)
+CREATE INDEX jidx_array ON testjsonb USING gin((j->'array'));
+-- gin index on expression not support for orca
+EXPLAIN SELECT count(*) from testjsonb  WHERE j->'array' ? 'bar';
+SELECT count(*) from testjsonb  WHERE j->'array' ? 'bar';
+-- type sensitive array exists - should return no rows (since "exists" only
+-- matches strings that are either object keys or array elements)
+-- gin index on expression not support for orca
+EXPLAIN SELECT count(*) from testjsonb  WHERE j->'array' ? '5'::text;
+SELECT count(*) from testjsonb  WHERE j->'array' ? '5'::text;
+-- However, a raw scalar is *contained* within the array
+EXPLAIN SELECT count(*) from testjsonb  WHERE j->'array' @> '5'::jsonb;
+SELECT count(*) from testjsonb  WHERE j->'array' @> '5'::jsonb;
+DROP INDEX jidx_array;
+
+--gin path opclass
+DROP INDEX jidx;
+CREATE INDEX jidx ON testjsonb USING gin (j jsonb_path_ops);
+
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"wait":null}';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"wait":"CC"}';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"wait":"CC", "public":true}';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"age":25}';
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{"age":25.0}';
+SELECT count(*) FROM testjsonb WHERE j @> '{"wait":null}';
+SELECT count(*) FROM testjsonb WHERE j @> '{"wait":"CC"}';
+SELECT count(*) FROM testjsonb WHERE j @> '{"wait":"CC", "public":true}';
+SELECT count(*) FROM testjsonb WHERE j @> '{"age":25}';
+SELECT count(*) FROM testjsonb WHERE j @> '{"age":25.0}';
+-- exercise GIN_SEARCH_MODE_ALL
+EXPLAIN SELECT count(*) FROM testjsonb WHERE j @> '{}';
+SELECT count(*) FROM testjsonb WHERE j @> '{}';
+DROP INDEX jidx;
+
+-- check some corner cases for indexed nested containment (bug #13756)
+create temp table nestjsonb (j jsonb);
+insert into nestjsonb (j) values ('{"a":[["b",{"x":1}],["b",{"x":2}]],"c":3}');
+insert into nestjsonb (j) values ('[[14,2,3]]');
+insert into nestjsonb (j) values ('[1,[14,2,3]]');
+create index on nestjsonb using gin(j jsonb_path_ops);
+
+explain select * from nestjsonb where j @> '{"a":[[{"x":2}]]}'::jsonb;
+explain select * from nestjsonb where j @> '{"c":3}';
+explain select * from nestjsonb where j @> '[[14]]';
+
+select * from nestjsonb where j @> '{"a":[[{"x":2}]]}'::jsonb;
+select * from nestjsonb where j @> '{"c":3}';
+select * from nestjsonb where j @> '[[14]]';
+
+CREATE INDEX wowidx ON test_tsvector USING gin (a);
+-- GIN only supports bitmapscan, so no need to test plain indexscan
+explain (costs off) SELECT count(*) FROM test_tsvector WHERE a @@ 'wr|qh';
+EXPLAIN SELECT count(*) FROM test_tsvector WHERE a @@ 'wr|qh';
+EXPLAIN SELECT count(*) FROM test_tsvector WHERE a @@ 'wr&qh';
+EXPLAIN SELECT count(*) FROM test_tsvector WHERE a @@ 'eq&yt';
+EXPLAIN SELECT count(*) FROM test_tsvector WHERE a @@ 'eq|yt';
+EXPLAIN SELECT count(*) FROM test_tsvector WHERE a @@ '(eq&yt)|(wr&qh)';
+EXPLAIN SELECT count(*) FROM test_tsvector WHERE a @@ '(eq|yt)&(wr|qh)';
+EXPLAIN SELECT count(*) FROM test_tsvector WHERE a @@ 'w:*|q:*';
+-- For orca, ScalarArrayOpExpr condition on index scan not supported
+EXPLAIN SELECT count(*) FROM test_tsvector WHERE a @@ any ('{wr,qh}');
+EXPLAIN SELECT count(*) FROM test_tsvector WHERE a @@ 'no_such_lexeme';
+EXPLAIN SELECT count(*) FROM test_tsvector WHERE a @@ '!no_such_lexeme';
+
+SELECT count(*) FROM test_tsvector WHERE a @@ 'wr|qh';
+SELECT count(*) FROM test_tsvector WHERE a @@ 'wr&qh';
+SELECT count(*) FROM test_tsvector WHERE a @@ 'eq&yt';
+SELECT count(*) FROM test_tsvector WHERE a @@ 'eq|yt';
+SELECT count(*) FROM test_tsvector WHERE a @@ '(eq&yt)|(wr&qh)';
+SELECT count(*) FROM test_tsvector WHERE a @@ '(eq|yt)&(wr|qh)';
+SELECT count(*) FROM test_tsvector WHERE a @@ 'w:*|q:*';
+-- For orca, ScalarArrayOpExpr condition on index scan not supported
+SELECT count(*) FROM test_tsvector WHERE a @@ any ('{wr,qh}');
+SELECT count(*) FROM test_tsvector WHERE a @@ 'no_such_lexeme';
+SELECT count(*) FROM test_tsvector WHERE a @@ '!no_such_lexeme';
+
+DROP INDEX wowidx;
+
+-- GIN index on complex array
+CREATE TABLE complex_array_table (complex_arr complex[]);
+CREATE INDEX ON complex_array_table USING gin (complex_arr);
+
+INSERT INTO complex_array_table VALUES (ARRAY[COMPLEX(1,3), COMPLEX(5,7)]);
+INSERT INTO complex_array_table VALUES (ARRAY[COMPLEX(2,4), COMPLEX(6,8)]);
+
+EXPLAIN SELECT * FROM complex_array_table WHERE complex_arr @> ARRAY[COMPLEX(2,4)];
+SELECT * FROM complex_array_table WHERE complex_arr @> ARRAY[COMPLEX(2,4)];
+-- with orca bitmap table scan off and table scan off, orca should fallback to
+-- planner to use bitmap index scan, as btree index plans are not supported with gin
+set optimizer_enable_tablescan=off;
+EXPLAIN SELECT * FROM complex_array_table WHERE complex_arr @> ARRAY[COMPLEX(2,4)];
+DROP TABLE complex_array_table;
+RESET enable_seqscan;
+RESET enable_bitmapscan;
+RESET optimizer_enable_tablescan;